Self's triage of Soul's first pass on hands/cut15-typed-state: - Soul views every ruling the cut_spec cites and makes one claim per ruling, beside the promises; Hands' self-report never bounds the pass. - Only Self closes a finding Fixed, after a Soul pass on the fix holds. Finding and verdict reads filter faculty Soul; faculty being attribution-only joins a new substrate-gaps table in campaign-state.md. - A fork before Hands' first commit ends the pass with the question alone; the blocked-spec check reads questions raised in a spec. - One rule for an organ that does not answer: missing tools stop, one retry on Unavailable (admit replays are safe), then a resumable stop that never sends the record as prose. - Survivor triage goes in deviations, one entry per survivor, because MutationRecord has no reason field (a recorded gap). - The target document is written and committed before the campaign and target cite it; Self's checklist names working_branch and target_doc. - Operator directions are rulings with no answers and the words verbatim. - One Rehydrate definition with a no-campaign / unknown-slug branch; per-brief recipes; Eyes reads the mind read-only; the postmortem template sources from the mind. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1.5 KiB
1.5 KiB
Postmortem template
Write the postmortem when the migration closes, from evidence: commits, the
campaign's mind, the prose map and transcripts. From the mind, take each cut's
record, each subject's history and the ledger (recipes in campaign-state.md);
cite document ids rather than restating them. Use an Eyes pass to gather facts
first. Report the delta and the scars, not a victory lap.
# <Migration> postmortem
## Summary
What changed, the scale (lines, dependencies, formats), and the state at
writing (landed, and what the operator still owes).
## Scope and invariants
The target in brief, what was excluded, and why the earlier attempt failed if
there was one.
## Timeline
One row per cut: dates, SHA ranges, Soul passes, and the notable finding.
## Structural delta
Estimate against actual, per repo, from the ledger recipe. What was deleted, added and parked.
## What Soul caught
Defects that green tests had passed, grouped by kind (concurrency, wire parity,
lifecycle and globals, untested rulings, split authority, spec errors). Say how
each would have failed.
## Operator corrections
Where the operator corrected the agents' understanding or overturned a plan, and
what context was missing that made the wrong answer look right.
## Incidents
Machine, tooling and process failures, with their cost and the rule each one
produced.
## What worked
Only what should be repeated, stated as practice.
## What to change in the pipeline
Concrete edits to the skill, each with its evidence.
## Open follow-ups
Item, owner, file, and why it can wait.